资源描述：
《基于一致性的分布式粒子滤波算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、AbstractDistributedParticleFilter(DPF)areapowerfulandversatileapproachtodecentralizedstateestimationinwirelesssensornetwork(WSN),theyareespeciallysuitedtolarge-scale,nonlinearandnon-Gaussiandistributedestimatedsystems.However,ithascertainlimitationsontheban
2、dwidthofcommunication,theresourceofnodes,thedynamicnetworktopologyandtheabilityofcommunicationandsoon.Inordertoreduceinfluencesofthelinksfailureorintermittentlinkthatexistinthenetworkforestimatingthetarget,thedeepstudyofdistributedparticlefilteralgorithmhas
3、agreatsignificance.Inviewofnon-linear,non-Gaussiantrackingapplicationinsensornetworks,thepaperproposesaconsensus/fusionbaseddistributedimplementationoftheparticlefilter(CF/DPF).Itrunstwoparticlefilters(PFs)ateachnode,theyarerespectivelylocalfilterwhichcomes
4、fromthedistributedimplementationoftheparticlefilterandthefusionfilterwhichcomputestheglobalfilteringdistribution.ThepaperapproximatestheproductofthelocalprobabilitydensityfunctionswithGaussiandistribution,andtheaverageconsensusalgorithmisusedtocalculatethep
5、arametersoftheGaussiandistribution,thenrealizethetargetstateestimation.thisalgorithmandthetrackingperformanceofcentralizedparticlefilteringwillbefinallycomparedWithMonteCarlosimulation,theyshowthatthealgorithmhasabetterfilteringperformance.Duetotheconvergen
6、cerateofconsensusisverycrucialintheabovealgorithm,sothepaperfocusesonprobability-basedtheweightoptimizationmethodofconsensus.Themethodintroducestheweightsoptimizationprobleminconsensusalgorithmsforspatiallycorrelatedrandomtopologies,itchoosestheconsensusmea
7、n-squareerror(MSE)convergencerateastheoptimizationcriterionandexpressesthisrateasafunctionofthelinkformationprobabilities,thelinkformationspatialcorrelationsandtheconsensusweights.BecausetheMSEconvergencerateisaconvex,non-smoothfunctionoftheweightforthesymm
8、etricrandomnetworks,thepapergivestheclosedformandsub-gradientalgorithmsolutiontosolvetheproblem.andtheoptimizationmethodiscomparedwithotherweightselectionmethodbythesimulation,resultsshowthattheweightd
